Wang Xingbo, Huang Renfei, Jin Zhihua, Fang Tianqing, Qu Huamin
IEEE Trans Vis Comput Graph. 2023 Oct 26;PP. doi: 10.1109/TVCG.2023.3327153.
Recently, large pretrained language models have achieved compelling performance on commonsense benchmarks. Nevertheless, it is unclear what commonsense knowledge the models learn and whether they solely exploit spurious patterns. Feature attributions are popular explainability techniques that identify important input concepts for model outputs. However, commonsense knowledge tends to be implicit and rarely explicitly presented in inputs. These methods cannot infer models' implicit reasoning over mentioned concepts. We present CommonsenseVIS, a visual explanatory system that utilizes external commonsense knowledge bases to contextualize model behavior for commonsense question-answering. Specifically, we extract relevant commonsense knowledge in inputs as references to align model behavior with human knowledge. Our system features multi-level visualization and interactive model probing and editing for different concepts and their underlying relations. Through a user study, we show that CommonsenseVIS helps NLP experts conduct a systematic and scalable visual analysis of models' relational reasoning over concepts in different situations.
最近，大型预训练语言模型在常识基准测试中取得了令人瞩目的成绩。然而，尚不清楚这些模型学到了哪些常识知识，以及它们是否只是利用了虚假模式。特征归因是一种流行的可解释性技术，用于识别模型输出的重要输入概念。然而，常识知识往往是隐含的，很少在输入中明确呈现。这些方法无法推断模型对上述概念的隐含推理。我们提出了常识可视化（CommonsenseVIS），这是一个视觉解释系统，它利用外部常识知识库将模型行为与常识问答的上下文联系起来。具体来说，我们在输入中提取相关的常识知识作为参考，使模型行为与人类知识保持一致。我们的系统具有针对不同概念及其潜在关系的多层次可视化以及交互式模型探测和编辑功能。通过一项用户研究，我们表明常识可视化有助于自然语言处理专家对模型在不同情况下对概念的关系推理进行系统且可扩展的视觉分析。
